Every construction accountant has left the same voicemail at least once this month. "Hey, it's me again, I just need Friday's hours and the concrete quantities so I can close the WIP schedule." Three days later the superintendent finally calls back from a truck somewhere on the highway, half the numbers are best guesses, and the job cost report goes out anyway because the deadline does not care that Friday was a mess. Ask a project accountant what the actual hard part of their job is and the honest answer is rarely the math. It is getting someone in a hard hat to pick up the phone.

That exact bottleneck, the office needing a straight answer from the field before the books can close, just got specific enough that a company built an AI agent whose job is to make that call, and investors paid $30 million to see it scale. Adaptive, a construction focused financial platform founded in 2021 by Matthew Calvano, Henry Bradlow and Francisco Enriquez, closed a $30 million Series B this week led by Tidemark, with existing investors Andreessen Horowitz, Emergence Capital, Pathlight, Definition and 3KVC returning. The round brings the company's total raised to $57 million and will fund expansion of what Adaptive calls its Project Accounting Agents, covering job costing, accounts payable, billing, work in progress, payments and compliance. More than 750 construction companies now run on the platform, ranging from $5 million to $1 billion in annual revenue, general contractors, specialty trades and real estate developers among them, spanning more than 10 accounting systems including QuickBooks, Sage Intacct, Foundation and Acumatica.

Calvano is direct about which problem he is actually solving, and it is not a spreadsheet problem. "The hardest part of construction accounting isn't the accounting," he said. "It's finding out what's actually happening on the job, and what that means for the numbers." That framing lines up with the named roles Adaptive gave its agents when it first unveiled them back in June: a WIP and Forecasting Analyst that keeps work in progress current as costs and revenue move, and, more to the point of that monthly voicemail, a Field Coordination Clerk built specifically to call superintendents and field crews for hours, project updates and missing information. The company is not automating the ledger entry, it is automating the phone call that has to happen before the ledger entry can be trusted.

Tidemark founder and managing partner Dave Yuan framed the bet in a category label that has been circulating in enterprise software this year: "Adaptive is one of the clearest examples we have found of a System of Action: it doesn't just help construction finance teams do the work, it increasingly does the work itself." That is a meaningfully different pitch than most construction fintech of the last decade, which mostly digitized the paper trail after the fact, better invoicing, better dashboards, better reports on money already spent. Adaptive's agents are working the process while it is still open, and the company says customers running them report cutting 70 to 80 percent of the time they used to spend on this specific slice of work.

The timing tracks with a broader shift the industry's own trade group is now measuring directly. The Associated General Contractors' 2026 Construction Hiring and Business Outlook found 61 percent of contractor respondents say their firms already use AI or plan to increase AI investment this year, up sharply from 44 percent in last year's survey. The breakdown is the interesting part: 45 percent are applying AI to office and administrative functions, more than the 23 percent using it for estimating or the 20 percent using it for design and preconstruction. The back office, unglamorous compared to jobsite robots and site cameras, is quietly turning out to be where agentic AI is landing first in construction, probably because it is the highest volume of repetitive judgment calls with the clearest paper trail to learn from.

There is an honest limit to sit with here too. An agent chasing job cost data is still only as good as what it can get out of a stubborn superintendent or a foreman who texts back the wrong crew count, and construction has spent a hundred years building workarounds for exactly that kind of messy field input. Raising a Series B is a sign investors believe the automation works well enough to keep scaling, not proof that every controller is ready to stop double checking the WIP number before it goes to a lender or a bonding company. The real test, the same one that shows up everywhere agentic AI touches this industry, is whether the accountant stops making the backup call once the agent's answer has earned enough trust to stand alone.

It is also a reminder that the data trail construction AI is racing to automate does not start at the job cost report, it starts at the bid. The same industry that is now comfortable letting an agent chase field hours and close a WIP schedule is the one that still runs most of its estimating and proposal work by hand, which is exactly the gap between winning the job and running its books that a tool like this makes more visible.

Sources